Interaction Forces and Adhesion of Supported Myelin Lipid Bilayers Modulated by Myelin Basic Protein Younjin Mina, Kai Kristiansena, Joan M

Home , Lipid bilayer

Interaction forces and adhesion of supported myelin lipid bilayers modulated by myelin basic protein Younjin Mina, Kai Kristiansena, Joan M. Boggsb,c, Cynthia Husteda, Joseph A. Zasadzinskia, and Jacob Israelachvilia,1

aDepartment of Chemical Engineering, University of California, Santa Barbara, CA 93106; bDepartment of Molecular Structure and Function, The Hospital for Sick Children, Toronto, ON, Canada M5G 1X8; and cDepartment of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, ON, Canada M5G 1L5

Contributed by Jacob Israelachvili, December 23, 2008 (sent for review September 26, 2008)

Force–distance measurements between supported lipid bilayers mimicking the cytoplasmic surface of myelin at various surface coverages of myelin basic protein (MBP) indicate that maximum adhesion and minimum cytoplasmic spacing occur when each negative lipid in the membrane can bind to a positive arginine or lysine group on MBP. At the optimal lipid/protein ratio, additional attractive forces are provided by hydrophobic, van der Waals, and weak dipolar interactions between zwitterionic groups on the lipids and MBP. When MBP is depleted, the adhesion decreases and the cytoplasmic space swells; when MBP is in excess, the bilayers swell even more. Excess MBP forms a weak gel between the surfaces, which collapses on compression. The organization and proper functioning of myelin can be understood in terms of physical noncovalent forces that are optimized at a particular combination of both the amounts of and ratio between the charged lipids and MBP. Thus loss of adhesion, possibly contrib- uting to demyelination, can be brought about by either an excess or deﬁcit of MBP or anionic lipids.

biomembrane adhesion ͉ lipid–protein interactions ͉ multiple sclerosis ͉ myelin membrane structure ͉ experimental allergic encephalomyelitis

he myelin sheath is a multilamellar membrane surrounding Fig. 1. The structure of the myelin sheath. The myelinated axon (A), myelin Tthe axons of neurons in both the central nervous system sheath (B), bilayer membranes (C), and phosphatidylethanolamine (PE) (D) are (CNS) and peripheral nervous system (PNS) (1) as shown in Fig. shown. Each bilayer of thickness DB is separated by cytoplasmic and extracel- 1 A and B. The myelin sheath consists of repeating units of lular water gaps of thicknesses DI and DO and effective protein thickness DP double bilayers separated by 3- to 4-nm-thick aqueous layers that occupied by the fraction of MBP constituting the cytoplasmic water gap. alternate between the cytoplasmic and extracellular faces of cell membranes (2) (Fig. 1C). Dehydrated myelin is unusual in that it D ϭ D ϩ D ϩ D ϩ D C is composed of 75–80% lipid and 20–25% protein by weight, where Tot [2 B I P O] (Fig. 1 ). Low capacitance ␧ compared with Ϸ50% of most other cell membranes (3) (Fig. 1 C necessitates a low value of myelin, which is promoted by the much and D). Multiple lipids make up the myelin sheath (Table 1), and lower dielectric constants of the lipid chains (␧hc Ϸ 2) and pro- each sheath, with its own distinct physical properties, contributes to teins (␧P ϭ 2.5–4.0) (6) compared with that of water (␧w Ϸ 80 the structure, adhesive stability, and possibly the pathogenesis of the for bulk water and ␧w Ϸ 40 for ‘‘interfacial’’ or ‘‘partially myelin membrane. The asymmetric distribution of lipid composi- trapped’’ water in thin films) (7). Increasing the thickness of tion on the cytoplasmic and extracellular faces likely also plays an the myelin sheath increases RO/RI, and tighter membrane important role (4). Myelin basic protein (MBP) constitutes 20–30% binding within the sheath decreases the water gaps, both of which of total protein by weight and is located only between the 2 decrease Cmyelin. For squid axon, a measured value of ␧myelin Ϸ 8.5 cytoplasmic faces, where it acts as an intermembrane adhesion has been reported (8). Inserting the following measured or protein. estimated values into Eq. 2: ␧P ϭ 2.5–4.0 (6), DB Ϸ 4.5 nm, The myelin sheath acts as an electrical insulator, forming a DO Ϸ 3 nm, DI Ϸ DP Ϸ 1 nm (assuming 50% of the cytoplasmic capacitor surrounding the axon, which allows for faster and more gap is protein) (1), and ␧w Ϸ 40 (7), we obtain ␧myelin Ϸ 13, which efficient conduction of nerve impulses than unmyelinated nerves is close to the measured value. (5). According to cable theory, the time to transmit a signal over Myelin dysfunctions vary from deterioration of signal trans- 1 2 a distance x is ␶ ϭ ⁄2RCmyelinx , where R is the resistance per unit duction to demyelinating diseases such as multiple sclerosis (MS) length, and Cmyelin is the capacitance between the axon and its (3). MS is characterized by the appearance of lesions, reflecting surroundings, which is given by (5) loss of membrane adhesion, swelling across the water gaps, vacuolization, vesiculation, and eventual disintegration of the 2␲␧0␧myelin Cmyelin ϭ per unit length, [1] log͑RO/RI͒ Author contributions: Y.M., K.K., J.A.Z., and J.I. designed research; Y.M. and K.K. performed research; J.M.B. and C.H. contributed new reagents/analytic tools; Y.M., K.K., and J.I. where RO and RI are the outer and inner radii (Fig. 1B), ␧0 is the permittivity of free space, and the mean dielectric constant of analyzed data; and Y.M., K.K., J.M.B., C.H., J.A.Z., and J.I. wrote the paper. the myelin sheath is The authors declare no conﬂict of interest. 1To whom correspondence should be addressed. E-mail: [email protected].

3154–3159 ͉ PNAS ͉ March 3, 2009 ͉ vol. 106 ͉ no. 9 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0813110106 Table 1. Lipid compositions of healthy and diseased bilayers calculated as mole % of lipid in the extracellular (EXT) and cytoplasmic (CYT) faces of myelin membranes Healthy EAE

EXT CYT TOT CYT EXT CYT TOT CYT Lipid type (i) EXT, Xi CYT, Xi Bilayer, Xi CYT, xi EXT, Xi CYT, Xi Bilayer, X CYT, xi * Cholesterol (CHOL) 22.4 10.6 32.9 31.6 25.8 12.1 37.9 37.4 Phosphatidylserine (PSϪ) 0.7 2.4 3.1 7.0 4.7 2.4 7.1 7.4 Hydroxylated cerebrosides (HCER) 13.9 0 13.9 0 14.7 0 14.7 0 Nonhydroxylated cerebrosides (NCER) 2.3 0 2.3 0 2.4 0 2.4 0 Cerebroside sulfatide (SCERϪ) 6.4 0 6.4 0 3.8 0 3.8 0 Sphingomyelin (SM Ϯ) 2.8 2.1 4.9 6.2 0.9 0.7 1.6 2.2 Phosphatidylcholine (PC Ϯ) 12.1 8.7 20.8 25.9 8.9 6.5 15.4 20.1 Phosphatidylethanolamine (PE Ϯ) 6.0 9.7 15.7 29.0 6.5 10.6 17.1 32.9

Total mole %, ⌺Xi and ⌺xi 66.5 33.5 100 100 67.8 32.2 100 100

TOT TOT EXT CYT Xi is the total mole fraction (percent of lipid of type i, so that ⌺iXi ϭ 100%. Xi and Xi are the total mole fractions of lipid of type i in the extracellular TOT EXT CYT TOT EXT CYT CYT and cytoplasmic faces, so that Xi ϭ Xi ϩ Xi , and ⌺iXi ϭ ⌺iXi ϩ ⌺iXi ϭ 100%. xi are the mole fractions of lipid of type i in the cytoplasmic faces CYT CYT CYT CYT only, so that ⌺ixi ϭ 100%. The xi and Xi values are related by xi ϭ 100Xi /⌺iXi . We may note that the total mole fraction of lipids in the extracellular faces EXT CYT ⌺iXi is typically twice that for the cytoplasmic lipids ⌺iXi (see bottom row). The ‘‘missing’’ volume is likely taken up by the MBP in the cytoplasmic spaces. Myelin membranes are from marmoset white matter (10). *These compositions were used in the experiments. myelin structure (9). It has been shown in previous work that in within the membranes (3) and the formation of compositionally experimental allergic encephalomyelitis (EAE) in the common distinct domains (17). marmoset—an established animal model for MS (10)—the lipid composition changes, in particular the ratios of charged to Results uncharged lipids (Table 1). In other studies, both the charged SFA Force–Distance Measurements. Fig. 2 shows the long- (A) and lipid and MBP isomer composition and the balance (relative short- (B) range forces measured between EAE cytoplasmic amounts) between the charged lipid and MBP isomer compo- bilayers in the absence and presence of various amounts of MBP. sition change in MS and EAE tissues (11, 12). We show here that Fig. 2A is a semilog plot showing the strongly repulsive forces on either an excess or a deficit of MBP relative to anionic lipids approach, and Fig. 2B is a linear plot showing the weak affects the interactions between myelin membranes across the (negative) adhesion forces† on separation as a function of bulk cytoplasmic space, although we note that whether swelling and MBP concentration. demyelination occur at both the extracellular and cytoplasmic To relate the measured adhesive forces to the surface coverage spaces during MS has not yet been established. rather than the bulk solution concentration of MBP,‡ we have Many attempts (10, 13, 14) have been made to experimentally separately measured the surface coverage at each concentration. observe the molecular interactions (forces) between myelin The same fringes-of-equal-chromatic-order (FECO) optical lipids and MBP. However, no techniques have quantified how Y technique (18) used to measure the distance between mica G O L D

electrostatic interactions (namely, double-layer, coulombic or O

surfaces can be used to determine the refractive index n(D) of I N B A L

ionic bonds, and salt bridges), which can be attractive or repul- S

the aqueous space between the bilayers, which can then be used A C I N S O I sive, and attractive van der Waals and hydrophobic interactions Y

to estimate the surface coverage of MBP. Because in a first T H A P

2 T act across the cytoplasmic and extracellular spaces of myelin to O I approximation, n (D) is a linear function of the volume fraction U B P ensure the integrity of its structure. The aim of this study was to M

␾i of each component i, n(D) is given by (19) O measure these forces, especially the adhesion forces, of recon- C 2 2 2 2 2 stituted myelin membranes at different lipid/protein ratios to n ͑D͒ ϭ ␾ini ϭ ␾wnw ϩ ␾BnB ϩ ␾MBPnMBP, [3] establish the correlations between membrane composition, ͸ structure, interaction forces, and ultimate function. where Α␾i ϭ 1 and subscripts w, B, and MBP denote water, To determine whether changes in the lipid or MBP concen- bilayer, and MBP, respectively. 2 tration or mole fraction affect the adhesion of the cytoplasmic In the absence of MBP (␾MBP ϭ 0), Eq. 3 reduces to n (D) ϭ 2 2 faces of the myelin sheath, we compared the forces between ␾wnw ϩ ␾BnB. In this case, ␾w ϭ Dw/D and ␾B ϭ (1 Ϫ ␾w) ϭ model bilayers of lipid mixtures that reflect the composition of 2DB/D, and we have: EAE myelin in the common marmoset and of normal, healthy myelin (10, 14). Model membranes with the composition of the 2DB n2 ͑D͒ ϭ n2 ϩ ͑n2 Ϫ n2 ͒, [4] cytoplasmic side of EAE myelin (see Table 1) were constructed w D B w on mica surfaces by Langmuir–Blodgett (LB) deposition, and the forces between the membranes were measured by using a surface where D ϭ 2DB ϩ Dw is the total mica–mica gap thickness and 2 2 2 forces apparatus (SFA) after exposure to various solution con- nw ϭ 1.333. Thus, by plotting n (D) Ϫ nw as a function of (nB Ϫ 2 centrations of MBP. Physical–chemical studies of the interaction nw)/D, we can determine 2DB and ␾B. This analysis gives 2DB Ϸ of MBP with myelin lipids have confirmed that the attraction 7.4 nm by using nB ϭ 1.47. between MBP and lipid is largely electrostatic (15): There are Ϸ20 Ϯ 3 anionic lipid molecules per molecule of MBP in healthy myelin (15). In addition, there is good evidence that the hydro- †Negative (adhesion) forces cannot be displayed on log plots. phobic segments that make up Ϸ25% of MBP (3, 15) interact ‡Most of the MBP injected into the solution ends up on the negatively charged surfaces; i.e., with the hydrophobic lipid chains either by partially penetrating there is an equilibrium distribution of MBP molecules between those molecules that are into the bilayers (16) or by expanding the bilayers (14, 15). These the bulk and those molecules adsorbed on the surfaces, which is strongly biased in the direction of the surfaces. From the concentration giving rise to monolayer coverage, the lipid–protein interactions also affect the intramembrane forces equilibrium afﬁnity or association constant is estimated to be 10Ϫ9 M, which is typical of that determine the fluidity and mobility of the lipids and proteins antibody–antigen and ligand–receptor bonds.

Min et al. PNAS ͉ March 3, 2009 ͉ vol. 106 ͉ no. 9 ͉ 3155 A 1000 0.156 μ (a) Approach 100 R

Mica )

Supported 2 bilayers Buffer 100 D (mN/m)

0.063μ Mica (mJ/m

R 10 R / π F , /2

MBP conc. F =

μ E adius , y

10 0.047μ 0.041 g e/R

c 0.031 r 1 o

0.016 Ener F 0-0.041μ 0

2DB Theoretical Debye length (slope) 1 0 10 20 30 40 50 60 70 80 Distance, D (nm)

2DB B 100 Separation 15

80 MBP conc. g/ml • ) 0.063 • g/ml 2 60 0.041 10 0.031 0.047 • g/ml 0.016 (mJ/m (mN/m) R 0 R / 40 F ,

5 /2 F =

20 E , adius y g e/R c

r 0 0 o Ener F No MBP Jumps out, -20 Max D Fadh j -5 0 5 10 15 20 Distance, D (nm)

Fig. 2. Normalized force–distance profiles with different amounts of MBP in solution. F(D)/R measured on approach (A) and on separation (B) between 2 EAE Fig. 3. Adsorbed density of MBP and adhesion forces with different amounts cytoplasmic myelin bilayers in the presence of various amounts of MBP (in ␮g of of MBP in solution. Calculations and measurements of injected MBP. (A) injectedMBPintothe80-mLincubatingMopsbuffersolution)atpH7.2and24 °C. Calculated surface coverage of MBP (⌫MBP) adsorbed between and bridging 2 D ϭ 0 corresponds to mica–mica contact, and DB represents 1 bilayer thickness as cytoplasmic EAE bilayers at increasing amounts, CMBP, of MBP injected into the defined in Fig. 1. The right axis shows the corresponding interaction energy, 80-mL Mops buffer solution at pH 7.2 and 24 °C. Full monolayer coverage of 2 E(D) ϭ F(D)/2␲R, per unit area between 2 planar bilayers, calculated according to MBP is shown by the line at 2 mg/m estimated from Eq. 6 by using ␾MBP ϭ 1 the Derjaguin approximation (23). The equilibrium separation between planar and D ϭ 3 nm, which corresponds to the thickness of C-shaped MBP (14). This bilayers corresponds to where F/R is a minimum; zero force in B is the equilibrium coverage agrees with the measured ⌫MBP at the maximum adhesion force and separation between 2 curved surfaces. Configuration of MBP in the bulk is minimum water-gap thickness (see B). (B) Adhesion forces measured, corre- primarily random coil (26, 37), which transforms into a C-shaped structure when sponding jump-out distances [Dj (which equals 2DB ϩ DI ϩ DP)], and water-gap it interacts with (bridges) 2 bilayers (38). Each curve corresponds to the second thicknesses [Dw (which equals DI ϩ DP)] at CMBP Յ 0.047 ␮g/mL]. Note that approach–separation cycle (compare Fig. 4). The numbers shown are the bulk Fadh/R first increases and then decreases to zero after showing a maximum at 2 MBP concentrations; the boxed concentrations refer only to the 4 curves in the ⌫MBP Ϸ 2 mg/m , whereas Dj and DW pass through a minimum close to where concentration range 0–0.041 ␮g/mL, which have matching colored data points. the adhesion is maximum. Black and white symbols refer to values measured The range of 2DB was obtained by subtracting the water layer thickness of Dw Ϸ on approach and separation (shown in Fig. 3A). The Inset in A is a sketch of the 2 nm from final ‘‘hard wall contact’’ distances. adsorbed bilayers showing how MBP most likely acts to couple the 2 bilayers together. Note that the absorption or surface coverage in A may not be linear ␾ Ͼ with the MBP concentration because MBP can grow as a gel layer on the In the presence of MBP ( MBP 0): surfaces (see Fig. 5). Most likely, the adsorption has 2 regimes, one to a full, 2 2 2 2 close-packed MBP monolayer (as in A) and the other corresponding to the n ͑D͒ ϭ ͑1 Ϫ ␾B Ϫ ␾MBP͒nw ϩ ␾BnB ϩ ␾MBPnMBP, growth of a gel layer. i.e.,

2 2 2 Fig. 3 shows the measured surface coverage of MBP (⌫MBP) n ͑D͒ Ϫ ␾BnB Ϫ ͑1 Ϫ ␾B͒nw ␾ ϭ . [5] between the 2 surfaces at the different bulk MBP concentrations MBP n2 Ϫ n2 MBP w tested. On correlating the data of Figs. 2 and 3, we find that as the bulk MBP concentration increases, the adhesion initially Assuming that the bilayer thickness, DB, and hence ␾B, do not increases, then decreases. Simultaneously, the equilibrium spac- change on addition of MBP and during compression, ␾MBP can be readily obtained from Eq. 5 by using measurements of n2(D) ing initially decreases, then increases. Significantly, maximum and also assuming that the refractive index of the protein is adhesion occurs when the equilibrium separation is the least (green and turquoise circles in Fig. 2), i.e., when the membranes known. Here, we assume nMBP Ϸ 1.55 (20, 21). Finally, the are in their most tightly packed configuration at a separation surface coverage of MBP trapped between the surfaces, ⌫MBP, Ϸ Ͼ ␮ can be estimated from ␾MBP as (18) of 12 nm. At MBP concentrations 0.041 g/mL, where the surface coverage exceeds some critical ‘‘saturation’’ value, 1 the adhesion disappears, and the magnitude and range of the ⌫ ϭ ␳ D␾ , [6] MBP 2 MBP MBP repulsion increases precipitously. At MBP concentrations below the saturation value (CMBP Ͻ 3 in terms of the known density ␳MBP (Ϸ1.38 g/cm ) (22). 0.041 ␮g/mL), quantitative analysis of the repulsive forces on

3156 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0813110106 Min et al. 2DB 1000 A C = 0.031 g/ml 120 120 MBP 100 CMBP = 0.156 µg/ml 80 10

60 800 a) 80 40 II P (MP

II , F/R (mN/m) I 20 Approach e 1

(mN/m) 0 R essur r (mN/m) / Separation 600 P

F -20 0 5 10 15 20 25 R , 40 /

Distance, D (nm) F 0.1 III , adius

dius 0 50 100 150 200 250

e/R 400 0 a

c Approach r Run 1 Out III Distance, D (nm) o e/R

F Run 2 Out

Jumps c I

r Separation Run 3 Out out, o

-40 Increasing F Run 4 Out Dj 200 number of compressions Run 5 Out 0 2 4 6 8 10 Water gap thickness, D (nm) W 0 0 50 100 150 200 250 300 350 B Distance, D (nm)

Fig. 5. Hysteretic forces in F/R vs. D runs measured in CMBP ϭ 0.156 ␮g/ml MBP solution. (Inset) When the force F/R between the 2 curved surfaces is converted to the pressure P between 2 planar surfaces (using the Derjaguin approximation to obtain E, then differentiating to get P ϭ dE/dD), and the calculated pressure vs. distance (proportional to the water gap volume) between 2 planar surfaces is plotted, the MBP behaves like a weak gel with a collapse pressure of P Ϸ 1 MPa. The collapse resembles a first-order phase transition because it has constant (horizontal) pressure regime over a large range of D (water Fig. 4. Effect of previous contacts and contact time on adhesion. (A) Adhe- I volume) corresponding to a collapse to Ϸ10% of the original volume occu- sion forces measured on separation in C 0.031 ␮g/mL MBP solution under mbp pied. The arrows indicate where the slopes of F/R and pressure change, and the the same conditions as in Fig. 2B. Successive force runs show a progressively corresponding letters represent the sequences from I–III. increasing adhesion and a concomitant decrease of the equilibrium (contact) separation brought about by a decreasing range of the stabilizing repulsive steric force. Water-gap thickness, DW, was obtained by subtracting 2DB Ϸ 7.4 D ϩ D nm from the final contact separations. The contact time in each cycle was Ͻ1 swelling also occurs in the absence of MBP, where I P min except in cycle 3, where the 2 surfaces were kept in contact for 30 min. increases by Ϸ2 nm (Fig. 2B). Note the linear relationship between the adhesion forces, Fadh/R, and the Figs. 4 and 5 give further details about the effects of the jump-out distances, Dj, except for the noticeable deviation in run 3. Inset number of approach–separation cycles (successive compressions shows the total force cycle measured on approach and separation (compare and decompressions), contact time, hysteresis effects in the Fig. 2) and the hysteresis in the forces (compare Fig. 5). A band of 2DB from forces, and the likely conformations of the MBP molecules 7.4–8.5 nm indicates a range of 2 bilayer thicknesses obtained by assuming a between the bilayers as a function of the coverage. We note, as constant bilayer thickness and by subtracting a typical water-layer thickness has been observed in many previous force measurements be- (Ϸ2 nm) from the final contact separation. (B) Possible model for the molec- tween biological samples (24), that the approach is more repul- ular rearrangements occurring on successive compressions that force the MBP Y sive than the separation. This effect is probably due to the G O

molecules to penetrate deeper into the bilayers and enhance their hydropho- L D O I initially rougher and less-correlated surfaces, which become N B

bic contacts (15, 28). Note that MBP penetrates deeper into the bilayers with A L S A

smoothed out, where attractive species diffuse toward each other on C each successive compression. This mechanism would explain the decreasing I N S O I * Y T range of the repulsion and the increasing adhesion. Note that DW becomes less contact to make the separation more attractive (or less repulsive). H A P T O I than DW after successive compressions whereas DB remains unchanged under The short-range forces measured on separation are therefore likely U B P the applied pressure range (Ͻ1 MPa). M to be closer to the ‘‘equilibrium’’ adhesion forces (14). O C At low MBP coverage, when CMBP ϭ 0.031 ␮g/mL (Fig. 4A), there was little hysteresis in the force vs. distance curves. Fig. 4A approach (Fig. 1A) shows the forces to be because of the further indicates that the adhesion force can be enhanced by electrostatic ‘‘double-layer’’ repulsion between the charged sur- repeated approach–separation cycles and also by increasing the faces: The forces are exponentially repulsive with a decay length equilibrium contact time between the 2 surfaces. This enhance- that is close to the theoretically expected Debye length of 0.8 nm ment shows that the lipids and proteins are laterally mobile and in 0.15 M NaNO3 solution (straight line in Fig. 2A). Thus, in ‘‘adaptable’’ and can diffuse and rearrange to find their optimum these near-physiological conditions, the forces are well-described configuration (14, 25). It should be noted that the adhesive force in by the Derjaguin–Landau–Verwey–Overbeek (DLVO) theory run 3 shows a noticeable deviation from the linear relationship (23) of attractive van der Waals and repulsive electrostatic shown by the straight line in Fig. 4A, supporting the idea that ‘‘double-layer’’ forces. However, in the presence of MBP, the increasing the contact (equilibration) time allows the binding sites depths of the adhesive wells are much deeper than can be on the mobile lipid and protein molecules to find each other (24). accounted for by van der Waals forces acting alone, indicating Another related observation is that the final contact and that the adhesion is also because of electrostatic bridging forces jump-out distances move farther in by repeated cycling. This [discrete ionic bonds between the negatively charged PS and decreasing trend in Dj suggests that the hydrophobic segments of lipids and the positively charged lysine or arginine amino acid MBP molecules penetrate into the lipid bilayers as shown (AA) groups in MBP] as illustrated in Fig. 4B and ref. 15. It is schematically in Fig. 4B, resulting in a thinning of the mem- also noticeable that the presence of MBP brings the 2 membranes. Because change occurs at constant lipid and protein branes closer together than in the absence of MBP, even as MBP coverage, it suggests a deeper penetration of the MBP into the gets sandwiched between the 2 bilayers. At higher MBP con- bilayers, i.e., the thinning is more of a smoothing of the mem- centrations, i.e., Ն0.047 ␮g/mL, there is excess MBP and, branes. Nevertheless, this process is accompanied by more water presumably, excess positive charge at the interface, which causes being forced out from the cytoplasmic gap, which further lowers the bilayers to repel each other electrostatically and the water the dielectric constant of myelin, which decreases the capaci- gap to swell (analogous to demyelination in vivo). But partial tance of the myelin sheath.

Min et al. PNAS ͉ March 3, 2009 ͉ vol. 106 ͉ no. 9 ͉ 3157 The forces are very different above the critical saturation To achieve optimum electrostatic conditions in the amounts and concentration of MBP. Fig. 5 shows the very large hysteresis charge ratio (or balance) of the positive MBP and negative lipids, observed at CMBP ϭ 0.156 ␮g/mL, well above the saturation both must be optimized with maximum adhesion occurring at concentration of 0.041 ␮g/mL. The corresponding pressure– maximum densities, where each negative charge on a lipid serine distance (or PV plot) between 2 planar surfaces has a shape that head group is bound to a positive arginine or lysine group on reveals a first-order phase transition of the MBP layer between MBP. Hydrophobic and other interactions also appear to be the surfaces, suggesting that it forms a weak and/or flexible gel involved both in the adsorption and intermembrane adhesion, when present in excess (26, 27). which are currently being investigated. Any compositional changes that lead to demyelination must also include changes to Analysis and Discussion of the Results. In the experiments, the mean the solution, such as pH and ionic changes or addition of solutes, area per lipid molecule is a Ϸ 0.4 nm2, measured during LB if these altered solutions lead to changes in the surface-charge depositions. This value appears to be lower than that expected density and/or dielectric constant that in turn affect the elec- for double-chain lipids and is attributed to the high amount of trostatic and hydrophobic interactions. We do not identify the cholesterol present in myelin (see Table 1). The fraction of primary causes that lead to these compositional changes, only negatively charged lipids is f Ϸ 0.074, which corresponds to a noting that these changes result in demyelination. negative surface-charge density of 0.4/0.074 ϭ 5.4 nm2 per unit Removed from optimum conditions, the bilayers repel and the charge eϪ. The estimate gives the excess number of positive water gap swells, which may correspond to the onset of demy- charges (arginine and lysine AAs) per 18.5-kDa MBP molecule elination. This swelling can be attributed to the charge neutral- to be Ϸ20 (15). Thus, for full-charge neutralization each protein ization, followed by charge overcompensation, which is also molecule should cover an area of 20 ϫ 5.4 Ϸ 108 nm2, which common in the flocculation of colloidal particles by polyelec- 23 corresponds to a surface coverage of ⌫MBP Ϸ [18,500/(6.02 ϫ 10 ϫ trolytes (30). Excess negative charge causes electrostatic swell- 108)] g/nm2 ϭ 0.28 mg/m2 per surface.§ This value, when compared ing, but excess positive charge, i.e., excess MBP, causes swelling 2 with the measured coverage of ⌫MBP Ϸ 2 mg/m obtained at through the formation of a weak gel of MBP (26), possibly with maximum adhesion (Fig. 3), suggests that other nonelectrostatic some lipid (11, 26, 31) that forces the bilayers apart.¶ This gel interactions also contribute to the adsorption. exhibits a phase transition on being compressed, which allows it Although the repulsive forces appear to follow the continuum to be collapsed when subjected to a compressive pressure of Ϸ1 theory of double-layer forces, the attractive forces do not MPa (10 atm). When the swelling pressures are not too great because at the measured jump-out distances the charges on the (e.g., 0.047 and 0.063 ␮g/mL in Fig. 2), the bilayers can be pressed lipid head groups and the charges on MBP must be very close to together again to almost the same separation, within 0–2 nm, as each other, if not actually in contact, so that continuum or mean the separation for adhesive, nonswelling systems, even from field theories can no longer be used. Therefore, regarding the distances as large as 80 nm (Fig. 2). maximum membrane–membrane adhesion force (as mediated These results indicate that swelling can have different causes: by MBP), assuming that at maximum adhesion every negative Excess or deficient charges on the lipid and/or MBP. For lipid is bound to a lysine or arginine group of MBP and that the example, arginine-deficient MBP has been found to be present adhesion force F/R is given by Ϫ2␲f␣␧/a (14, 24), where again f Ϸ in some MS cases (32). To counteract the swelling, drainage of 0.074 is the fraction of negatively charged lipids, a Ϸ 0.4 nm2 is the water gap by using nontoxic hydrophilic polymers like the mean area per lipid molecule, ␣ Ϸ 1 is the fraction of positive polyethylene glycol (PEG) (33) has been successfully used to charges on MBP that are bound to the membranes, and ␧ Ϸ 2kT treat spinal-cord injuries in animals (34). PEG induces a high is the Coulomb energy for an ionic bond in water (24), then the osmotic pressure (attractive depletion force) that forces the maximum F/R is calculated to be Ϸ9 mN/m. This value is lower bilayers together again, perhaps collapsing the gel layer pro- than the maximum adhesion force of Ϸ19 mN/m measured at duced by unbound MBP. It also appears that there needs to be CMBP ϭ 0.031 ␮g/ml (Fig. 3), suggesting that other interactions an as-yet-unknown mechanism to regulate production of MBP to such as van der Waals, hydrophobic (15, 28), hydrogen-bonding match the existing surface charge in the cytoplasmic space. The interactions (29), and weak dipolar interactions (28) also con- insight into cytoplastic myelin bilayer interactions and swelling tribute to the attractive forces, which is also consistent with the presented here suggests guidelines for assessing and possibly structural-thickness and refractive-index changes (adsorbed arresting or reversing myelin swelling. MBP) observed. We note that hydrophobic- and hydrogen- bonding interactions have been shown to be important in the Materials and Methods interactions of P0 glycoprotein in PNS myelin (29). Our previous work (10) identiﬁed the changes in overall lipid composition Finally, the MBP undergoes conformational changes when it between normal (control) and EAE myelin in the white matter of the common marmoset. Previous force measurements (14) by using the mean total lipid adsorbs to the lipid bilayer (25). In the presence of excess MBP, a composition of normal and EAE myelin were made in the absence and pres- phase transition is observed across highly swollen bilayers, suggest- ence of an undetermined quantity of MBP. However, as shown in refs. 2 and ing the formation of a weak, dilute, but extended gel-like structure 4, the cytoplasmic and extracellular faces of the myelin bilayer have very of MBP rather than a simple electrostatic repulsion (26, 27). different compositions. We used the distributions of the lipids determined by We conclude that MBP maintains the structure and stability Inouye and Kirschner (2, 4) to calculate the composition of the cytoplasmic and of the cytoplasmic region of the myelin sheath by acting as an extracellular monolayers of the control and EAE membranes (Table 1), given electrostatic ‘‘glue’’ holding the negatively charged bilayers the overall lipid compositions of normal and EAE myelin measured in our together through its positively charged amino acids groups and previous work (10). The asymmetry in lipid composition between the extra- that hydrophobic and other attractive (adhesive) interactions are cellular and cytoplasmic monolayers was believed to be the same for both control and EAE (4, 15). also involved. Maximum adhesion coincides with minimum PhosphatidylserineϪ (porcine brain PSϪ), sphingomyelin (porcine brain water-gap thickness of the cytoplasmic space, which further SM), phosphatidylcholine (porcine brain PC), phosphatidylethanolamine (por- implies minimum dielectric constant (or capacitance) of the cine brain PE), and cholesterol (ovine wool), all of purity Ͼ99%, were pur- myelin and, therefore, maximum transmission of nerve signals. chased from Avanti Polar Lipids and stored in chloroform until used. The major

§This estimate assumes that each protein is adsorbed fully extended on each surface as in ¶Repulsion is also enhanced by polyunsaturated lipids through their increased repulsive Fig 2A. If the MBP bends back when bridging 2 surfaces as in Fig. 3A, this does not change undulation forces. Interestingly, EAE lipids have a greater degree of polyunsaturation the calculated coverage per surface. than in healthy bilayers, which may also play a role in promoting demyelination (10).

3158 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0813110106 Min et al. chain lengths of the 3 major phospholipids (PC, PE, and PSϪ) are 16:0, 18:0, at a surface pressure of 35 mN/m from the solid phase. The second monolayer 18:1, and 20:4. Dipalmitoylphosphatidylethanolamine (DPPE), sodium nitrate, of 23.0 mass % (37.4 mole %) cholesterol, 9.6 mass % (7.4 mole %) PSϪ, 2.6 calcium nitrate, and morpholinepropanesulfonic acid (Mops) sodium salt were mass % (2.2 mole %) SM, 24.3 mass % (20.1 mole %) PC, and 39.0 mass % (32.9 purchased from Sigma-Aldrich. MBP was isolated from bovine-brain white mole %) PE was made in a 11:5:4 (vol/vol) solution of hexane/chloroform/ matter as described (35). The MBP used was unfractionated, heterogeneous, ethanol. After the second ‘‘myelin’’ monolayer deposition at 30 mN/m onto bovine MBP. the first DPPE monolayer, one of the bilayer-covered surfaces was transferred An SFA 2000 was used for the force measurements (23, 24). The crossed- into the SFA chamber in buffer solution. The other surface was incubated with cylinder geometry of the surfaces (each cylinder of radius R) is locally equiv- different amounts of MBP in buffer solution for 30 min, then rinsed twice with pure buffer solution to remove MBP molecules from the bulk and those loosely alent to a sphere of radius R interacting with a flat surface or 2 spheres of adsorbed to the surface before transferring into the SFA chamber, producing radius 2R. A simple geometric transformation, called the Derjaguin approxi- an ‘‘asymmetric’’ bilayer system consisting of a pure lipid bilayer facing an mation (23), converts the force–distance curve [F(D)] measured between the MBP-covered bilayer. 2 curved surfaces to the energy–distance curve, E(D) ϭ F(D)/2␲R, or pressure– The forces F as a function of distance D were measured between the 2 distance curve, P(D) ϭ dE(D)/dD, between 2 flat (planar) surfaces (23), corre- curved mica-supported bilayers in solutions of varying MBP concentrations as sponding to the interaction geometry of myelin membranes. previously described (14). The optical technique used in the measurements The lipid bilayers were deposited on the mica surfaces by using conven- gave the surface radius R (Ϸ2 cm), surface separation D (to Ϸ0.1 nm), and tional Langmuir–Blodgett deposition (36). The composition used was the independently the refractive indices of the various components (water, bi- cytoplasmic side of EAE membranes, because MBP is found exclusively on the layer, protein) between the surfaces and hence an estimate of the in situ cytoplasmic side of the membrane. Monolayers were spread from solvent onto amounts of these components during an interaction. a pH7.2 Mops buffer (150-mN sodium nitrate/10 mM Mops sodium salt/2 mM calcium nitrate) and compressed to the desired surface pressure after solvent ACKNOWLEDGMENTS. This work was supported by National Institutes of evaporation. As the first layer, DPPE was deposited onto the mica substrates Health Grant R01 GM076709.

1. Raine CS (1984) in Myelin, ed Morell P (Plenum, New York), pp 1–41. 21. Fisk AA (1950) The thicknesses of hemoglobin and bovine serum albumin molecules as 2. Inouye H, Kirschner DA (1988) Membrane interactions in nerve myelin. 1. Determina- unimolecular layers adsorbed onto films of barium stearate. Proc Natl Acad Sci USA tion of surface-charge from effects of pH and ionic-strength on period. Biophys J 36:518–523. 53:235–245. 22. Smith R (1982) Self-association of myelin basic protein: Enhancement by detergents 3. Williams KA, Deber CM (1993) The structure and function of central nervous system and lipids. Biochemistry 21:2697–2701. myelin. Crit Rev Clin Lab Sci 30:29–64. 23. Israelachvili J (1992) in Intermolecular and Surface Forces (Academic, London), pp 169–250. 4. Inouye H, Kirschner DA (1988) Membrane interactions in nerve myelin. 2. Determina- 24. Leckband D, Israelachvili J (2001) Intermolecular forces in biology. Q Rev Biophys tion of surface-charge from biochemical data. Biophys J 53:247–260. 34:105–267. 5. Jeans J (1966) in The Mathematical Theory of Electricity and Magnetism (Cambridge 25. Mueller H, Butt HJ, Bamberg E (2000) Adsorption of membrane-associated proteins to Univ Press, Cambridge, UK), pp 332–335. lipid bilayers studied with an atomic force microscope: Myelin basic protein and 6. Gilson MK, Honig BH (1986) The dielectric constant of a folded protein. Biopolymers cytochrome c. J Phys Chem B 104:4552–4559. 25:2097–2119. 26. Smith R (1992) The basic protein of CNS myelin: Its structure and ligand-binding. 7. Cherepanov DA, Feniouk BA, Junge W, Mulkidjanian AY (2003) Low dielectric permit- J Neurochem 59:1589–1608. tivity of water at the membrane interface: Effect on the energy coupling mechanism 27. Rapis E (2007) On the nonequilibrium phase transition in protein. Tech Phys 52:787–792. in biological membranes. Biophys J 85:1307–1316. 28. Roux M, Nezil FA, Monck M, Bloom M (1994) Fragmentation of phospholipid bilayers 8. Northrop RB (2001) in Introduction to Dynamic Modeling of Neuro-Sensory Systems by meylin basic protein. Biochemistry 33:307–311. (CRC, Boca Raton, FL), pp 11–12. 29. Luo X, et al. (2007) Cytoplasmic domain of zebrafish myelin protein zero: Adhesive role 9. Hafler DA (2004) Multiple sclerosis. J Clin Invest 113:788–794. depends on beta-conformation. Biophys J 93:3515–3528. 10. Ohler B, et al. (2004) Role of lipid interactions in autoimmune demyelination. Biochim 30. Bouyer F, Robben A, Yu WL, Borkovec M (2001) Aggregation of colloidal particles in the Biophys Acta 1688:10–17. presence of oppositely charged polyelectrolytes: Effect of surface charge heteroge- 11. Riccio P, Fasano A, Borenshtein N, Bleve-Zacheo T, Kirschner DA (2000) Multilamellar neities. Langmuir 17:5225–5231. Y

packing of myelin modeled by lipid-bound MBP. J Neurosci Res 59:513–521. 31. Haas H, et al. (2004) Small angle X-ray scattering from lipid-bound myelin basic protein G O L

12. Husted C (2006) Structural insight into the role of myelin basic protein in multiple in solution. Biophys J 86:455–460. D O I N B sclerosis. Proc Natl Acad Sci USA 103:4339–4340. 32. Wood DD, Bilbao JM, Oconnors P, Moscarello MA (1996) Acute multiple sclerosis A L S A C I

13. Jo EJ, Boggs JM (1995) Aggregation of acidic lipid vesicles by myelin basic protein: (Marburg type) is associated with developmentally immature myelin basic protein. N S O I Y

Dependence on potassium concentration. Biochemistry 34:13705–13716. Ann Neurol 40:18–24. T H A P T O

14. Hu YF, et al. (2004) Synergistic interactions of lipids and myelin basic protein. Proc Natl 33. Kuhl TL, Berman AD, Hui SW, Israelachvili J (1998) Part 2. Crossover from depletion I U B P

Acad Sci USA 101:13466–13471. attraction to adsorption: Polyethylene glycol induced electrostatic repulsion between M O

15. Boggs JM, Moscarello MA, Papahadjopoulos D (1982) in Lipid–Protein Interactions, eds lipid bilayers. Macromolecules 31:8258–8263. C Jost PC, Grifﬁth OH (Wiley, New York), pp 1–51. 34. Borgens RB, Bohnert D (2001) Rapid recovery from spinal cord injury after subcutane- 16. Bates IR, Feix JB, Boggs JM, Harauz G (2004) An immunodominant epitope of myelin ously administered polyethylene glycol. J Neurosci Res 66:1179–1186. basic protein is an amphipathic alpha-helix. J Biol Chem 279:5757–5764. 35. Cheifetz S, Moscarello MA (1985) Effect of bovine basic-protein charge microhetero- 17. Albersdorfer A, Feder T, Sackmann E (1997) Adhesion-induced domain formation by geneity on protein-induced aggregation of unilamellar vesicles containing a mixture interplay of long-range repulsion and short-range attraction force: A model mem- of acidic and neutral phospholipids. Biochemistry 24:1909–1914. brane study. Biophys J 73:245–257. 36. Zasadzinski JA, Viswanathan R, Madsen L, Garnaes J, Schwartz DK (1994) Langmuir– 18. Raviv U, et al. (2002) Properties and interactions of physigrafted end-functionalized Blodgett ﬁlms. Science 263:1726–1733. poly(ethylene glycol) layers. Langmuir 18:7482–7495. 37. Krigbaum WR, Hsu TS (1975) Molecular-conformation of bovine A-1 basic-protein, a 19. Israelachvili J, Sammut RA, Snyder AW (1976) Birefringence and dichroism of photo- coiling macromolecule in aqueous-solution. Biochemistry 14:2542–2546. receptors. Vision Res 16:47–52. 38. Keniry MA, Smith R (1979) Circular dichroic analysis of the secondary structure of 20. Barer R, Tkaczyk S (1954) Refractive index of concentrated protein solutions. Nature myelin basic protein and derived peptides bound to detergents and to lipid vesicles. 173:821–822. Biochim Biophys Acta 578:381–391.

Min et al. PNAS ͉ March 3, 2009 ͉ vol. 106 ͉ no. 9 ͉ 3159